When I checked my node reference status screen this morning (on the Linux/SunJava node), I saw that my list of refs was down to:
Number of node references: 15 Contacted node references: 13 Backed off node references: 1 Total Trials: 10128 Total Successes: 3882 It was in the upper 30s last night when the node had been running for only a couple hours. My OpenBSD/Kaffe node has a much healthier list: Number of node references: 49 Contacted node references: 46 Backed off node references: 7 Total Trials: 14291 Total Successes: 6423 And it's been running for the same length of time, give or take a few minutes. (In fact, that list is SO healthy that I think I'm going to save it!) It seems to me that there is some sort of random factor here -- a node that has a bit of bad luck reaching its peers will go into a downward spiral in which it can't reach other nodes, so it deletes references to them, so it has fewer references from which to choose, so it can't reach other nodes, .... But a node that has a bit better luck in reaching its brethren will prosper. I've observed this before during the 0.4 days. Of course this isn't a rigorous scientific experiment. These two nodes are *not* configured identically (and shouldn't be, because they're running on very different hardware -- the clock speed alone is more than three times as high on the OpenBSD box). And I haven't attempted to repeat this experiment. They're both running the same build, though - CVS branch rel-0-5-1 built last night (2002-10-31) at 18:26 US/Eastern. -- Greg Wooledge | "Truth belongs to everybody." [EMAIL PROTECTED] | - The Red Hot Chili Peppers http://wooledge.org/~greg/ |
msg01882/pgp00000.pgp
Description: PGP signature